Am 16.04.2013 um 17:35 schrieb Arnau: > Hi Reuti, > > So, from your answer, submitting a 8 slots jobs show suspend my low prio job?
Correct, or you have to adjust setting of the "subordinate_list". > > job-ID prior name user state submit/start at queue > slots ja-task-ID > ----------------------------------------------------------------------------------------------------------------- > 470056 1.05292 low abria r 04/16/2013 17:30:48 > low-el6@aracne13. 8 > 470057 0.06218 high abria qw 04/16/2013 17:31:00 > 8 > > now it says: > > cannot run in PE "smp" because it only offers 0 slots > > Thanks for your answer, > Arnau > > > 2013/4/16 Reuti <[email protected]> > Hi, > > Am 16.04.2013 um 14:49 schrieb Arnau Bria: > > > I wanted to start playing with subordination. I've started with > > Queuewise + suspension. > > > > so I've created a couple of queues: high and low, and a hgrp of one node > > @preempt. > > > > > > > > [...] > > qname high-el6 > > hostlist @preempt > > slots 8 > > subordinate_list low-el6=8 > > low-el6 will be suspended when all 8 slots are filled according to this etup. > You submitted only a serial job. > > -- Reuti > > > > [...] > > > > > > [...] > > qname low-el6 > > hostlist @preempt > > slots 8 > > subordinate_list NONE > > [...] > > > > # qconf -shgrp @preempt > > group_name @preempt > > hostlist aracne13 > > > > # qconf -se aracne13 > > hostname aracne13 > > load_scaling NONE > > complex_values slots=8,virtual_free=60G You allow only 8 slots per machine, hence 8+1 wasn't able to start like 8+8 now. It's necessary to remove this limit here (i.e. don't specify slots at all) -- Reuti > > load_values arch=linux-x64,num_proc=8,mem_total=48257.140625M, \ > > [...] > > np_load_long=0.000000 > > processors 8 > > > > # qconf -sc|grep slots > > slots s INT <= YES YES 1 > > 1000 > > > > > > # qstat -f|egrep 'low|high' > > [email protected] BIP 0/0/8 0.04 linux-x64 > > [email protected] BIP 0/8/8 0.04 linux-x64 > > > > So, I submit a job to low-el6, wait until it start and send a job to > > high queue, but the low queue job is not suspended: > > > > $ echo sleep 1000 |qsub -N low -q low-el6 -pe smp 8 > > > > $ qstat -q low-el6,high-el6 > > job-ID prior name user state submit/start at queue > > slots ja-task-ID > > ----------------------------------------------------------------------------------------------------------------- > > 467796 1.05292 low abria r 04/16/2013 14:43:04 > > low-el6@aracne13 8 > > > > > > $ echo sleep 100 |qsub -N high -q high-el6 > > > > $ qstat -q low-el6,high-el6 > > job-ID prior name user state submit/start at queue > > slots ja-task-ID > > ----------------------------------------------------------------------------------------------------------------- > > 467796 1.05292 low abria r 04/16/2013 14:43:04 > > low-el6@aracne13 8 > > 467828 0.00000 high abria qw 04/16/2013 14:44:21 > > 1 > > > > $ qstat -j 467828 > > [...] > > (-l h_rt=21600,virtual_free=12G) cannot run at > > host "aracne13.linux.crg.es" because it offers only hc:slots=0.000000 > > [...] > > > > > > > > I've read man and soc about subordination, and seems quite easy, but > > I'm missing something and I can't figure out what it is... > > > > Someone with more experience could give a hand? > > > > TIA, > > Arnau > > _______________________________________________ > > users mailing list > > [email protected] > > https://gridengine.org/mailman/listinfo/users > > _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
